behavioral biases in new product forecasting

12
International Journal of Forecasting 3 (1987) 393-404 North-Holland 393 BEHAVIORAL BIASES IN NEW PRODUCT FORECASTING * Tyzoon T. TYEBJEE Santa Clara University, Santa Clara, CA 95053, USA Abstract. The new product planning process generates an upward bias in the forecast of a product’s performance. Three sources of such bias are discussed: (1) the post-decision audit bias reflects a regression-to-the-mean phenomenon since only those products that are forecasted to do well, including those with the most upward biased forecasts, are brought to market; (2) the advocacy bias reflects the tendency of product planners to champion their project by overpromising on forecasts; (3) the optimism bias results from the act of participating in planning activities. Two role-playing experiments found that persons who were more deeply involved in a planning exercise were more optimistic about the outcome of the plan than those who were less involved. A third role-playing experiment demonstrated that one reason for the optimism bias is that during the planning process the illusion of control over the environment leads the planner to change assumptions about uncontrollable events which are likely to affect the outcome. Keywords: Bias, Forecast error, Judgmental forecasting, New products, Optimism, Planning bias. 1. Introduction Most recent textbooks on new product marketing stimulate an interest in the topic by citing the rate of new product failure. Crawford (1979), in a review of 31 studies of new product introduction, estimates the failure rate to lie in the 30% to 40% range. A more recent study reports that 2.5% of new industrial products that make it to market are judged by management to be failures [Cooper (1982)]. However, there is no evidence that the rate of product failure is decreasing. Booz Allen and Hamilton (1982) surveyed 700 U.S. businesses in 1981 and found that a third of the products commercialized are considered to be failures and that this rate was essentially unchanged from their earlier study in 1968. Since failure is typically defined in these studies as a product not living up to some performance goal (e.g., sales, market share, or profitability), a product can be designated a failure either because it did not perform well or because management’s expectations were unrealistically high. This paper concentrates exclusively on the second of these two perspectives as to why a product is viewed by management as a failure. This, of course, does not imply that high expectations cause failure, but rather that failure is judged not merely in terms of performance but also in terms of expectations. * This research was completed while the author was on sabbatical at INSEAD, Fontainebleau, France. The support of INSEAD is gratefully acknowledged. The author is also indebted to Meir Statman for stimulating his interest in this topic and to Scott Armstrong and Shelby McIntyre for their constructive criticisms of an earlier draft. The Gunning Fog Index for this paper is about 14. 0169.2070/87/$3.50 0 1987, Elsevier Science Publishers B.V. (North-Holland)

Upload: tyzoon-t-tyebjee

Post on 22-Nov-2016

215 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Behavioral biases in new product forecasting

International Journal of Forecasting 3 (1987) 393-404

North-Holland 393

BEHAVIORAL BIASES IN NEW PRODUCT FORECASTING *

Tyzoon T. TYEBJEE

Santa Clara University, Santa Clara, CA 95053, USA

Abstract. The new product planning process generates an upward bias in the forecast of a product’s performance. Three sources of such bias are discussed: (1) the post-decision audit bias reflects a regression-to-the-mean phenomenon since only those products that are forecasted to do well, including those with the most upward biased forecasts, are brought to market; (2) the advocacy bias reflects the tendency of product planners to champion their project by overpromising on forecasts; (3) the optimism bias results from the act of participating in planning activities. Two role-playing experiments found that persons who were more deeply involved in a planning exercise were more optimistic about the outcome of the plan than those who were less involved. A third role-playing experiment demonstrated that one reason for the optimism bias is that during the planning process the illusion of control over the environment leads the planner to change assumptions about uncontrollable events which are likely to affect the outcome.

Keywords: Bias, Forecast error, Judgmental forecasting, New products, Optimism, Planning bias.

1. Introduction

Most recent textbooks on new product marketing stimulate an interest in the topic by citing the rate of new product failure. Crawford (1979), in a review of 31 studies of new product introduction, estimates the failure rate to lie in the 30% to 40% range. A more recent study reports that 2.5% of new industrial products that make it to market are judged by management to be failures [Cooper (1982)]. However, there is no evidence that the rate of product failure is decreasing. Booz Allen and Hamilton (1982) surveyed 700 U.S. businesses in 1981 and found that a third of the products commercialized are considered to be failures and that this rate was essentially unchanged from their earlier study in 1968. Since failure is typically defined in these studies as a product not living up to some performance goal (e.g., sales, market share, or profitability), a product can be designated a failure either because it did not perform well or because management’s expectations were unrealistically high. This paper concentrates exclusively on the second of these two perspectives as to why a product is viewed by management as a failure. This, of course, does not imply that high expectations cause failure, but rather that failure is judged not merely in terms of performance but also in terms of expectations.

* This research was completed while the author was on sabbatical at INSEAD, Fontainebleau, France. The support of INSEAD is gratefully acknowledged. The author is also indebted to Meir Statman for stimulating his interest in this topic

and to Scott Armstrong and Shelby McIntyre for their constructive criticisms of an earlier draft. The Gunning Fog Index

for this paper is about 14.

0169.2070/87/$3.50 0 1987, Elsevier Science Publishers B.V. (North-Holland)

Page 2: Behavioral biases in new product forecasting

394 Z T. Tyebjee / Behacioral buses in new product forecasting

Studies on the accuracy of sales forecast show them to be upwardly biased. Tull’s (1967) survey of new product introductions by 24 companies found actual sales to fall short of forecast for two thirds of the new products. Ferber (1953), Hastay (1954) and Theil (1961) report similar results. Studies of new product development costs find cost overruns to be prevalent across a wide variety of industries [see Mansfield et al. (1971), Marshall and Meckling (1962). Norris (1971), Peck and Scherer (1962) Thomas (1970/1971)].

The most recent studies that compare forecasted revenues and costs against actual outcomes are in the transportation industry [Gordon (1986), Webber (1976), Wachs (1985) Wachs and Ortner (1979)]. The experience of rapid transit projects in San Francisco, Washington, D.C., Atlanta, Baltimore, and Miami all show a pattern of ridership below forecast and both capital and operating costs to be above forecast. In the case of San Francisco’s BART system, 1976 ridership was only 51% of the forecast for that year in a 1962 report advocating that the system be built. Capital costs for construction and rolling stock were 240% of forecast, and operating costs were 516% of forecast. The Baltimore system is expected to run at half the forecasted patronage used to justify its construction. Incredibly, the advocates of each transit system have ignored the dismal performance of the forecasts used to justify previous systems. Los Angeles is about to build a multi-billion dollar rail transit system. Over the years, as inflation has driven the estimate of capital costs of the proposed subway from around $2 billion to $3.5 billion, forecasts of daily patronage have followed suit, increasing from 180,000 to 300,000. These revisions in ridership forecast seem to have been motivated by a decision to maintain the same proportionate relationship between the benefits and costs rather than because of any real change in the planned system and its environment.

Transportation projects, like utilities projects and defense projects, are probably more capital intensive and more influenced by political factors than most new product projects undertaken in the private sector. However, it can be argued that the planning process within any organizational setting is subject to the same kind of influences that result in biased forecasts of sales and revenues. New product forecasts are not merely the result of a technical forecasting process but rather are the result of a complex behavioral process influenced by the values, goals, and roles of many members of the organization [for the perspective of several disciplines on behavioral biases in planning, see Barnes (1984) Hogarth and Makridakis (1981) Kahneman and Tversky (1979) and Taylor (1982)]. Even the most technically objective forecasting models require many subjective assumptions. The choice of a particular data transformation or functional form. seemingly a technical detail, can change the model’s forecast from increase to decrease, contraction to growth, loss to gain [Wachs (1982)].

This paper concentrates on the behavioral biases that influence the subjective component of forecasts made during a planning process. In particular, three behavioral influences that result in inflated sales forecasts and deflated cost forecasts in new product proposals and a consequent tendency towards disappointment in actual profit performance are discussed. These three behavioral biases in new product forecasts, namely the post-decision audit bias, the advocacy bias, and the optimism bias, are considered in the following sections.

2. Post-decision audit bias

Why is the actual performance of new products, on the average, worse than expected? One possible explanation is that the upward bias in sales forecasts (downward bias in the case of cost forecasts) is a result of the sample of products for which the accuracy of forecasts can be examined.

A simple, albeit contrived, example helps explain the post-decision audit bias. Consider the case of a company faced with the possibility of introducing two products, A and B. A 15% market share is

Page 3: Behavioral biases in new product forecasting

T. T Tyebjee / Behavioral biases in new product forecastq 395

required in each case to achieve the minimum expected return and warrant a GO decision. Market research studies that have taken all the necessary steps to ensure a representative sample and unbiased measurement show that market share estimates for the two products A and B are 12% and 22%, respectively. In fact, unknown to the marketer, the true estimate of share in the population is 17% for both products, and the deviation in observed estimates is due to sampling error. In the typical company, product A’s launch will be aborted because it does not meet performance criteria for minimum acceptable return. Product B will be launched and will achieve the 17% market share that was the true population value. When the company evaluates its forecasting system, it can do so only against the results achieved by the product introduced and it will find that its forecast was

upwardly biased, i.e., 22% vs. actual share of 17%. However, in the aggregate over the two products A and B there was no systematic bias.

The observed bias is a result of the fact that the forecasting method can be audited only against products that were actually introduced and that these products will over-represent the cases where the forecasts overestimated reality. This post-decision audit bias is not a result of poor forecasts but rather of rational decision rules based on point estimates that are subject to estimation error. The bias results from a regression-to-the-mean phenomenon in which low values never have the chance to

regress upwards.

3. Advocacy bias

In most organizational contexts, the task of preparing new product plans and accompanying forecasts and the task of evaluating these plans from the point of view of resource allocation are not in the hands of the same person. For example, a product manager who develops a plan for new products must submit pro forma sales, cost, and profit projections, along with a supporting budget, for review and evaluation by the group product manager. The latter must decide whether to allocate resources to this proposal in the broader context of other competing proposals. In other organiza- tions, the proposal development may be done by a venture team and the evaluation by a new product committee, but the essential fact of the separation of planning and resource allocation remains.

Let us hypothesize that persons evaluating proposals view new product plans with a degree of skepticism and informally discount the sales and cost estimates proposed by the planner. If we do this, we may see a corresponding tendency on the part of the planner to anticipate the discounts that will be applied to the forecast and, accordingly, to inflate the forecast of sales and deflate the estimates of the costs to develop, manufacture, and support the product.

Cyert, March and Starbuck (1961) gave a group of MBA students ten new product proposals, each characterized by a low and a high estimate of the proposed product’s unit cost. The students were asked to play the role of analyst and, for each proposal, select a point estimate within the range estimate provided. Another group of students were given exactly the same range estimates for the ten proposals but these were identified as sales forecasts rather than cost estimates. The study found that the point estimates were skewed toward the high end of the range in the cost version but toward the low end in the sales version. On the basis of this result, the study concludes that people evaluating forecasts expect them to be biased toward a rosy picture (lower costs, higher sales). Accordingly, the evaluators discount the forecasts.

Statman and Tyebjee (1985) replicated the Cyert, March and Starbuck study in order to examine the influence of the evaluator’s work experience on their discounting behavior, The sample included undergraduate business majors, evening MBA students (the majority of whom worked full-time), and managers enrolled in an executive development program, The sample included persons with a wide range in years of work experience. The instructions given to the respondents were as follows:

Page 4: Behavioral biases in new product forecasting

396 T. T. Tyebjee / Behacioral biases rn new product forecasring

Assume you are the vice president of product development in a company. You are evaluating eight new product proposals. For each proposal you have a forecast of

(a) Research and development costs, (b) Average annual sales in the first five years after product introduction.

You have made it a practice to ask two persons in research and development, A and B, in whom you have equal confidence, to give you independent forecasts of research and develop- ment costs. You also ask two persons in marketing, X and Y, in whom you have equal confidence, to give you independent forecasts of sales. These forecasts are given below for the eight proposals. In order to make a financial analysis of each proposed product, you must make your forecast for research and development cost and sales, in the case of each proposal. Your forecast may be based on the estimates provided to you, although the actual figure you choose does not have to be identical to any of the forecasts given to you.

The sample was subdivided, on the basis of years of work experience, into three groups of approximately 50 respondents each. One group had fewer than three years of full time work experience, the second group has three to five years, and the third group over five years. Each respondent’s revision of the cost and sales forecasts was scaled from 0 to 1. A score of 0 represents personal estimates at the low end of the range provided, averaged across the eight proposals evaluated. Similarly, scores of 1.0 and 0.5 represent the high end of the range and mid-point respectively. The results of the study are shown in exhibit 1. The results corroborate Cyert, March and Starbuck’s finding that those who evaluate forecasts will adjust sales forecasts downward and cost forecasts upward. In addition, the results in exhibit 1 demonstrate that the extent to which the forecast is discounted depends on the number of years of work experience of the person evaluating the forecast. Those with a significant amount of work experience demonstrate higher skepticism about the forecast. Studies in non-laboratory settings corroborate this result. Rubinstein and Schroder’s (1977) study in an R&D facility found that the higher-ranked managers assigned lower subjective probabilities to the technical success of projects.

One interpretation of the results of the studies described above is that work experience makes a person more aware of the bias in cost and sales forecasts (an alternative interpretation is that work

experience, and its correlate age, make decision-makers more conservative). Accordingly, the evalua- tor discounts the forecast. The contention in this paper is that those who wish to champion a new product are aware of the discounting rules that will be applied by those evaluating the product proposal and the accompanying forecast. In anticipation of the discount, they inflate their forecast in order to ensure an endorsement of the proposal by the evaluator in spite of the latter’s discounts.

In other words, the person who proposes a new product anticipates the evaluator’s discount rule and inflates expectations vis-a-vis the new product. As a result, the product’s plan includes a forecast that exceeds even the planner’s true expectations. Once these biased forecasts become documented in the plan, they become the bench mark against which performance will be judged. This gives rise to an advocacy bias when the forecast is compared to the actual outcome after the product’s launch.

Neither the Cyert, March and Starbuck study nor the Statman and Tyebjee study conclusively demonstrate that planners exhibit an advocacy bias. They merely demonstrate that evaluators discount forecasts made in a proposed plan. However, it is not surprising that those with work experience would be aware of this behavior by evaluators and will thus behave in an anticipatory manner. In a broader sense, from the planners’ perspective a forecast is not made merely to reduce uncertainty about the future but rather to influence the resource allocation process to support the proposal. In this way, forecasts take on a political meaning. This is not to portray the planner or

Page 5: Behavioral biases in new product forecasting

T. T. Tyebjee / Behuvioral biases in new product forecasting 397

1.0 1

0.9 -

0.8 -

I- 0.7 -

P t2

0.6 -

b 0.5 -- t- 5 z 0.4 -

3, ( 0.3 -

0.2 -

0.1 -

DISCOUNTS INFLATE FORECABTS

SALES FORECAST

DISCOUNTS DEFLATE

<3 yearn 3-5 yeam xi years

WORK EXPERIENCE

Exh. 1. Evaluation adjustments for advocacy biases.

forecaster as a Machiavellian individual who falsifies information. Rather, the advocacy bias acknowledges that forecasting is impossible without assumptions and planning is impossible without goals. Taken together, we see that the assumptions may well be self-serving to the goals.

It is tempting to argue that the advocacy bias can be eliminated by using forecasters who are independent of those championing a proposal. In fact, even independent forecasters are influenced by allegiances implicit in their assignment.

In addition to commitment to a body of tools and techniques, the forecaster must also have loyalty or responsibi~ty to the agency which he or she serves, either as an employee or as a consultant. The employee wishes to advance and wants to be considered both competent and cooperative by his or her superiors. The consultant wishes to be considered for future contracts . . . . Forecasts often require so many assumptions that there is leeway to allow the forecaster to satisfy both organizational goals and technical criteria. Indeed, if he or she has become a ‘ team player’ and has internalized the goals of the agency, there may not even appear to be a conflict between the two loyalties. In cases where the forecaster is aware of the conflict, and where reasonable technical judgment may deliver Forecasts which the agency would rather hear about, the forecaster faces the problem of choosing between advocacy and objectivity. The rewards for advocacy are clear, while even the criteria for judging objectivity are ambiguous [Wachs (1982, p. 567)].

Page 6: Behavioral biases in new product forecasting

T T. TJlehjee / Behavioral biases in new> product forecusting

4. Optimism bias

On the basis of case studies of planning in several firms, Cyert, Dill and March (1958, p. 339) report that ‘expectations were by no means independent of such things as hopes, wishes and the internal bargaining needs of sub-units in the organization . . both conscious and unconscious bias in expectations is introduced’. In the previous section we discussed the conscious bias introduced by planners in forecasts in order to improve their bargaining leverage for resources. We called this type of bias an advocacy bias. In this section, we explore the provocative possibility that the very act of planning introduces an optimism bias into the forecasts upon which the plan is based. Unlike the advocacy bias, the optimism bias is not a result of any conscious factor.

Optimistic biases result from the need for decision-makers to act. Doubts must be cast aside so that the decision process can move forward [Balderston (1987)]. The planner is constantly flip-flop- ping from an analytic mode to an action mode and back again. The uncertainties encountered in the analytic mode become inhibitors in the action mode. The way these uncertainties are resolved or discarded helps explain the optimism bias.

Matlin and Stang (1978) reviewed a large number of experiments in cognitive and social psychology to persuasively demonstrate the ‘Pollyanna principle’. People show an inherent bias toward favoring positive stimuli in many of the processes inherent in the act of planning, such as information (stimuli) selection, perception, learning, retention, and retrieval. Also, the planner may be affected by wishful thinking and the illusion of control over an uncertain environment [see Hogarth and Makridakis (1981)]. For example, the higher the value of an outcome to an individual, the higher will be the probability assigned to achieving that outcome [Morlock (1967), Slavic (1966)]. Moreover, engaging in cognitive activity prior to entering a situation in which the outcome is

determined by factors outside a person’s control results in the illusion that the uncontrollable factors can be influenced to one’s advantage [Langer (1975)].

Miller, Kets de Vries and Toulouse (1982) report that business executives view environmental constraints as malleable. These executives scored higher on internal locus of control relative to the general population. Among the sample of 24 executives, those with the highest internal locus of control pursued the most product-market innovation, undertook greater risks, and tended to lead rather than follow competition. Larwood and Whittaker (1977) asked students to estimate the sales potential for a product they were to introduce as managers of a hypothetical firm. The product was a me-too version of those offered by five existing competitors, all of whom had equal shares of the market. The pro forma projections of the students showed that they expected their entry to grow faster than the market and assume market share leadership within three years.

Finally, it is possible that the advocacy bias described in the previous section becomes internalized by the planner. As a result, the planner begins to accept the consciously biased forecast as being true expectations. In other words, advocacy may induce optimism.

The forecasting process is often an integral part of the planning process. Planning is a cognitive activity subject to many of the biases that accompany such activity. Among these biases are selective receptivity to supportive information, perceptual distortion of information, over-weighting prior beliefs relative to contrary information, selective retention of favorable experiences, and an illusion of control over a potentially contrary environment. Consequently, we hypothesize that engaging in planning activity causes optimism in the planner about what the plan can achieve.

This hypothesis is a direct consequence of many of the biases in forecasting and planning cataloged by Hogarth and Makridakis (1981). These biases and their relationship to the optimism

bias are given in the table. Three experiments, described below, were designed to test the hypothesis that planning induces

optimism. In all these experiments, the subjects were students enrolled in an evening MBA program.

Page 7: Behavioral biases in new product forecasting

T. T Tyebjee / Behavioral biases in new product forecasting 399

Type of bias Relationship to optimism bias

Availability

Selective

perception

Social pressures

Overconfidence

Wishful thinking

Illusion of

control

Success/failure

attributions

Organizations tend to publicize new product success stories more than failures. Hence, the

frequency of success is overestimated.

Anticipation of what one expects to see biases what one does see; people seek information

consistent with their hypotheses; people downplay or disregard conflicting evidence

Individuals will not voice opinions that threaten any consensus.

The uncertainty in a prediction is considered to be lower than that warranted by the available

facts.

People’s preferences for outcomes of events affect their assessment of the events.

Activity concerning an uncertain outcome results in feelings of control over the uncertain event.

Tendency to attribute success to one’s skill and failure to chance: past failures are discounted

The experiments were administered by the author during normal class hours. Subjects were led to believe they were engaged in a case study in which they were to assume the perspective of a member of a planning team assigned to produce a forecast for a new product this team was developing. The average work experience in the samples was five years; and 78% of the subjects were working full-time, typically in technical, management, or administrative jobs.

Experiment 1

Two groups of 31 and 32 subjects (randomly assigned) were asked to estimate the probability that a manufacturer of a new type of car would achieve its sales objective. For the first group, the marketing expenditures on consumer advertising, consumer sales promotions, dealer promotions, and fleet sales promotions were specified by the experimenter. Subjects in the second group were free to choose, that is, to plan the level of each of the marketing expenditures but constrained not to exceed the levels that had been prescribed for the first group. This constraint was necessary to ensure that the hypothesized optimism of the second group did not arise from their choice of larger marketing budgets than the amount allowed to the first group. We would hypothesize that the act of planning, as represented by the freedom to set the budgets, would result in higher estimated probabilities of meeting sales goals relative to no involvement in planning.

Experiment 2

Eighty subjects were randomly assigned to 1.5 groups of five to six persons. Each group was given the task of devising a marketing plan for a new product concept (home security device) provided to them. Within each group, only three of the members (randomly assigned) were allowed to engage in the planning discussion. The remaining members were asked to act as silent observers of the planning process. At the end of the 45minute planning session the planners within each group were asked to provide a consensus estimate of the market penetration (percentage of the U.S. households) the new product would achieve after five years from introduction. Similarly, the observers within each group separately provided their consensus forecast. Again, we would hypothesize that planners who,

Page 8: Behavioral biases in new product forecasting

400 T. T Tyebjee / Behavioral biases in new product forecasting

relative to observers, are more involved in the planning activity would be more optimistic about the penetration the product would achieve.

Experiment 3

Fifty-four subjects were randomly assigned to 11 groups. Each group was given 45 minutes to develop a marketing plan for a new product. The product, described in detail, was an electronic beeper that a parent could use to monitor a child’s safety. It would sound an alarm if the child wandered away, was abducted, or fell into water. The description of the product was such that the size of its market would be affected by five environmental factors, all of which were uncontrollable by the marketer of the product. These factors were described as (1) an increase in the birth rate; (2) an increase in the size of the U.S. population under 10 years of age; (3) an increase in kidnapping crimes; (4) an increase in households with swimming pools and hot-tubs; and (5) development of competitive products. Each subject rated the likelihood of occurrence of each of the five factors on a 7-point scale (1 = very unlikely, 7 = very likely). The subjects made these ratings twice, once before beginning the planning session and then again at the end of the planning session. We would expect that the act of planning would induce the illusion of control whereby planners upgrade the likelihood of favorable events and downgrade the likelihood of unfavorable events, even though these euents are

outside their sphere of influence.

Results

The results are displayed in table 1. Overall, the direction of the results support the hypothesis that planning induces optimism. In experiment 1, the subjects who were allowed the discretion to decide the level of the marketing budget predicted a higher probability of meeting sales objectives than those who were not given this option, even though the former group was constrained to selecting a spending level at or below that imposed on the latter group (p < 0.01). In experiment 2, those persons who actively engaged in preparing a marketing plan for a new product forecasted a higher market penetration for that product than those who were passive observers of the very same planning process ( p -c 0.10).

Experiment 3 differs from the first two in two ways. First, it uses pre-planning and post-planning measures of optimism and consequently looks at the longitudinal effect in the planning process. Second, whereas the first two experiments merely establish whether planning induces optimism, the third experiment tests whether this optimism arises because of an illusion of control. The results indicate that, after engaging in planning activities, the planners considered the uncontrollable environment to be more favorable than before they began the planning task. Of the five factors affecting the market potential of the product described in the experiment, planners upgraded the likelihood of more favorable conditions in three cases, namely, birth rate (p < O.lO), size of population below 10 years of age ( p < 0.05), and freedom from competition ( p < 0.001). The change in a fourth factor, threat of child abduction, was in the expected direction though not significant. All these factors, of course, cannot be determined by the manager’s plan, and yet, after having participated in the planning activity, the subjects seemed to adjust their expectations as if they could be. For one of the factors, the number of homes with pools or hot-tubs, the shift in expectations was opposite to the direction hypothesized.

Of the seven results across the three experiments (see table l), six were in the predicted direction and, of these six, all but one were statistically significant. Rosenthal (1978) has suggested several

Page 9: Behavioral biases in new product forecasting

T. T Tyebjee / Behavioral biases in new product forecasirng 401

Table 1

The effect of planning on optimism.

Experiment Dependent variable Treatment conditions

Low High

involvement involvement

in planning in planning

t-test of difference in means

f-value one-tail

p-value

(1)

(2)

(3)

Estimated probability

of meeting sales goals

Estimated 5 year

market penetration (W)

Likelihood of:

(a) Increase in birth rate

(b) Increase in

population below

10 of years age

(c) Increase in child

kidnapping

(d) Increase in homes

with swimming pools

and hot tubs

(e) Competitive entry

within 2 years

0.54 0.67 2.58 0.006

(n = 31) (n = 32)

7.0

(n =15)

Prior to

planning

(n = 54)

3.89

4.46

4.33

4.22

5.63

10.2

(n =15)

After

planning

(n =54)

4.13

4.69

4.39

4.13

5.04

1.54 0.073

1.56 0.062

2.20 0.016

0.50 0.308

-0.82 0.792

-3.12 0.001

methods by which independent studies testing a common directional hypothesis can be combined into one meta-analysis. For each of three of the methods proposed by Rosenthal (adding logs, adding probabilities and adding t’s for the seven tests in table l), the null hypothesis that planning does not induce optimism is rejected (p < 0.001).

5. Managerial implications

The post-decision audit bias is an important issue for reaffirming management’s faith in seemingly inadequate forecasting systems. In fact, Harrison and March (1984) show that even random and unbiased forecast errors result from a tendency toward post-decision disappointment when managers make the defensible choice of the alternative with the highest estimated value. Recognition of the fact that, on the average, we can expect to be disappointed in the decisions we make on the basis of forecasts should make managers shift the bench mark against which the performance of a forecasting system is judged. To put it slightly differently, managers should focus on whether forecasting systems help in the rational pursuit of maximizing profits as opposed to the questionable goal of minimizing later disappointment, that is, post-decision audit bias.

Forecast biases due to optimism and advocacy, unless appropriately deflated by the manager evaluating such forecasts, would result in the allocation of resources to new products that in fact have marginal or poor prospects. Such biases are more likely to occur when those preparing the new product proposal are either not responsible for implementing the plan or not accountable for any shortfall in the achievement of these forecasts. This means that the pay-off/loss function faced by

Page 10: Behavioral biases in new product forecasting

402 T. T Tyebjee / Behaworul binses in new product forecastmg

the planner is asymmetric in that the rewards associated with having the plan approved are not adequately balanced by the negative consequences of not delivering on the expectations generated in the plan [McIntyre and Statman (1982)].

One would think that reward structures that include stronger disincentives for not meeting forecasts would reduce both advocacy biases and optimism biases, and result in more realistic forecasts. However, in practice this may be difficult to achieve for several reasons. First, the success of a new product is often not known for a considerable period of time after its launch. By this time, those responsible for the preparation of the forecast may no longer be associated with the product or company. Second, the results achieved by a new product are subject to a wide variety of uncontrolla- ble events and it may be impossible to assign accountability in the event that the product does not live up to expectations. In fact, studies of how management accounts for success versus failure shows a predictable tendency toward attributing success to the actions of management and attributing failure to environmental factors beyond the manager’s control [Bettman and Weitz (1983), Salancik and Meindl (1984) and Staw. McKechnie and Puffer (1983)].

Quite aside from the issue of whether planners can be induced to reduce advocacy and optimism biases, it is not clear whether the firm should do so. Once a plan is approved, the forecast upon which the plan is based, in effect, becomes the goal that the plan intends to achieve. Any attempts by those evaluating a plan to formally adjust downward the expectations generated in a plan results in goals that are imposed from outside and have a lower level of aspiration. Studies of goal setting and performance, on the other hand, suggest that performance is highest when goals are self-generated and more challenging in terms of the difficulty of attainment [Latham and Wexley (1981, pp. 190-191) Latham and Locke (1984, p. 21)].

Ultimately, of course, the effectiveness of new product development programs is determined not so much by the accuracy of forecasts generated in the planning process but rather by the commitment of those who manage the product to fulfill the goals and expectations that are an integral part of the forecast. Schwenk (1986) takes the position that planning biases should in fact be encouraged because they build commitment. In this view, commitment to a plan is built by inducing overconfidence and an illusion of managerial control over the environment. The dilemma of those who must evaluate planning and development activities and at the same time motivate those activities is how to inject realism into economic decisions and yet be supportive of the optimism and enthusiasm that are critical motivational ingredients in a successful product development effort.

6. Summary

This paper has discussed several reasons why the actual performance of a new product can be expected to be worse than that predicted in its marketing plan. These reasons lie in three types of biases; namely the post-decision audit bias, the advocacy bias, and the optimism bias. The forecasting literature is surprisingly sparse in evaluating the bias in judgmental forecasts. Though there is no lack in this literature of studies that have assessed the accuracy of subjective forecasts against actual outcomes. these studies almost universally focus on absolute or squared error measures of accuracy, and thereby obscure the direction of the deviation, that is, whether they forecasted too much or too little. It is hoped that this paper will stimulate future research to consider bias as well as error in subjective expectations.

The paper treats forecasting as an integral part of the planning process. This view has provided the opportunity to integrate several theoretical perspectives such as organizational theory, decision theory, cognitive economics, and social psychology as they bear upon marketing planning. This integration should bring forth more questions about marketing planning processes. Clearly, several

Page 11: Behavioral biases in new product forecasting

T. T. Tyebjee / Behavroral biases in new product forecasfrng 403

questions remain unasked and unanswered in this paper. Are marketing managers more optimistic than technical people? Is optimism in planning peculiar to certain cultures? What is the effect of group decision-making on optimism in forecasts? These and other issues establish an agenda for

future research.

References

Balderston, Frederick E., 1987, Facade and self-deception in the financial firm, California Management Review 29, 101-111.

Barnes, James H., 1984, Cognitive biases and their impact on strategic planning, Strategic Management Journal 5, 129-137.

Bettman, James R. and Barton A. Weitz, 1983. Attributions in the boardroom: Causal reasoning in corporate annual reports,

Administrative Science Quarterly 28, 165-183.

Booz Allen and Hamilton, 1982, New product management for the 1980s (Booz Allen and Hamilton, New York).

Cooper, Robert G., 1982, New product success in industrial firms, Industrial Marketing Management 11, 215-223.

Crawford, C. Merle, 1979, New product failure rates: Facts and fallacies, Research Management 22, 9-13.

Cyert, Richard M., W.R. Dill, and James G. March, 1958, The role of expectations in business decision-making, Administra-

tive Science Quarterly 3, 307-340.

Cyert, Richard M., James G. March and William H. Starbuck, 1961. Two experiments on bias and conflict in organizational

estimation, Management Science 7, 254-264.

Ferber, Robert, 1953, Measuring the accuracy and structure of businessmen’s expectations, Journal of the American Statistical

Association 48, 385-413.

Gordon. Peter, 1986, Transit foolishness in Los Angeles, Working paper, School of Urban and Regional Planning, University

of Southern California, Los Angeles.

Harrison, J. Richard and James G. March, 1984, Decision making and post-decision surprises, Administrative Science

Quarterly 27, 26-42.

Hastay. M., 1954. The Dun & Bradstreet surveys of businessmen’s expectations, Proceedings of the Business and Economic

Statistics Section of the American Statistical Association, 93-123.

Hogarth, Robin M. and Spyros Makridakis, 1981, Forecasting and planning: An evaluation, Management Science 27,

115-138.

Kahneman, Daniel and Amos Tversky, 1979, Intuitive prediction: Biases and corrective procedures. TIMS Studies in

Management Sciences 12, 313-327.

Langer, E.J., 1975, The illusion of control, Journal of Personality and Social Psychology 32, 311-328.

Larwood, Laurie and William Whittaker, 1977, Managerial myopia: Self-serving biases in organizational planning, Journal of

Applied Psychology 62, 194-198.

Latham, Gary P. and Edwin A. Locke, 1984, Goal setting: A motivational technique that works (Prentice-Hall, Englewood

Cliffs, NJ).

Latham, Gary P. and Kenneth N. Wexley, 1981, Increasing productivity through performance appraisal (Addison-Wesley,

Reading, MA).

Mansfield, Edwin et al.. 1971, Research and innovation in the modem corporation (Norton, New York).

Marshall, A. and W. Meckling, 1962, Predictability of the costs, time and success of development, in: National Bureau of

Economic Research, The rate and direction of inventive activity: Economic and social factors (Princeton University Press,

Princeton, NJ).

Matlin, Margaret and David Stang, 1978, The Pollyanna principle: selectivity in language, memory and thought (Schenkman, Cambridge, MA).

McIntyre, Shelby and Meir Statman, 1982, Managing the risk of new product development, Business Horizons 25, 51-55.

Miller, Danny, Manfred F. Kets de Vries and Jean Marie Toulouse, 1982, Top executive locus of control and its relationship

to strategy-making, structure and environment, Academy of Management Journal 25, 237-253.

Morlock, H., 1967, The effect of outcome desirability on information required for decisions, Behavioral Sciences 12, 296-300.

Norris, K.P., 1971, The accuracy of project cost and duration estimates in industrial R&D, R&D Management 2, 25-36.

Peck, M.J. and F.M. Scherer, 1962, The weapons acquisition process: An economic analysis (Harvard University Press, Cambridge, MA).

Rosenthal, Robert, 1978, Combining results of independent studies, Psychological Bulletin 85, 185-193.

Rubinstein, Albert H. and Hans-Horst Schroder, 1977, Managerial differences in assessing probabilities of technical success for R&D projects, Management Science 24, 137-148.

Salancik. Gerald R. and James R. Meindl, 1984, Corporate attributions as strategic illusions of management control. Administrative Science Quarterly 29, 238-254.

Page 12: Behavioral biases in new product forecasting

404 T. T Tyebjee / Behavioral biases in new product forecastmg

Schwenk, Charles, 1986, Information, cognitive biases and commitment to a course of action, Academy of Management Review 11, 298-310.

Slavic. P., 1966, Value as a determiner of subjective probability, IEEE Transactions in Human Factors in Electronics HFE-7, 22-28.

Statman, Meir and Tyzoon T. Tyebjee, 1985, Optimistic capital budgeting forecasts: An experiment, Financial Management 14, 27-33.

Staw, Barry M., Pamela I. McKechnie and Sheila M. Puffer. 1983, The justification of organizational performance, Administrative Science Quarterly 28, 582-600.

Taylor, Ronald N., 1982, Organizational and behavioral aspects of forecasting, in: Spyros Makridakis and Stephen C. Wheelwright, eds., The handbook of forecasting (John Wiley, New York).

Theil, Henry, 1961, Economic forecasts and policy (North-Holland, Amsterdam).

Thomas, H., 1970/1971, Some evidence on the accuracy of forecasts in R&D projects, R&D Management 1, 55-69.

Thomas, H., 1970/1971, The debiasing of forecasts in research and development, R&D Management 1, 119-123.

Tull, Donald S., 1967. The relationship of actual and predicted sales and profits in new product introductions, Journal of Business 40, 233-250.

Wachs. Martin, 1985, Management vs. political perspectives on transit policymaking, Journal of Planning Education and Research 4, 1399147.

Wachs. Martin, 1982, Ethical dilemmas in forecasting for public policy, Public Management Forum 42, 562-567.

Wachs, Martin and James Ortner, 1979, Capital grants and recurrent subsidies: A dilemma in American transportation policy, Transportation 8, 3-19.

Webber, Melvin M., 1976, The BART experience: What have we learned?, Monograph 26, Institute of Transportation Studies, University of California, Berkeley.

Bibliography: Tyzoon TYEBJEE is Associate Professor of Marketing at the Leavey School of Business at Santa Clara University. His research interests are fairly eclectic, including international joint ventures, high technology start-ups and venture capital, in addition to topics in marketing strategy and new product planning.